Notes from active research.
Long-form writing on applied AI, methodology, prototyping, infrastructure, evaluation, and the practical edge of enterprise AI adoption. Written by the GET team for the technical decision-makers we work with.
All published notes
Why prototypes fail when evaluation is skipped
A working demo and a deployable prototype are not the same artifact. The gap is where most pilots die quietly. Why evaluation belongs in the build phase, not after it.
Designing private systems for sensitive environments
When public cloud assumptions break, the architecture shifts. Notes from on-premise and air-gapped research — what changes, what doesn't, and where the surprises usually are.
The difference between a demo and a deployable prototype
Demos optimize for the path of least resistance. Deployable prototypes optimize for the path of most resilience. The cost of conflating them — especially in AI work — and how to scope around it.
Where AI actually creates leverage inside an organization
AI adoption fails when it starts with tools and looks for problems. A short framework for finding the parts of an operation where intelligent systems create real, measurable leverage.
Building on-premise AI systems for sensitive data
On-prem is not just cloud-without-the-cloud. The constraints reshape the architecture, the deployment model, and the kind of model you can run in the first place.
How to benchmark early-stage technical systems
Early benchmarks are mostly wrong, and that's fine — what matters is whether they're wrong in a structured way. A short framework for benchmarking before the system is finished.
Separating real capability from market hype
New AI models ship faster than evaluations of them. A working approach for assessing whether a frontier capability is actually production-ready, or whether the demo is doing all the work.
When "we can probably build this" isn't a yes
Most things are technically possible. The relevant question is whether they're worth building, at what cost, and with what residual risk. Why feasibility is a budget question, not just a technical one.
Have a technical challenge worth investigating?
Bring us the problem. We will help determine what is possible, what is practical, and what should be built next.
Response within two business days · NDAs available when required